Source-primed Multi-turn Conversation Helps Large Language Models Translate Documents

2025-03-13

AI Summary

This work addresses the challenges of omission errors and contextual incoherence in document-level machine translation with large language models (LLMs). To this end, we propose a source-guided, multi-turn dialogue-style translation framework. The method segments the source document and performs iterative, turn-wise translation, reusing prior key-value (KV) caches to preserve cross-segment context. Crucially, we introduce a novel “source-primed” prompting mechanism—preloading source segment information before target generation—which enables cross-segment consistency modeling and efficient inference without any fine-tuning. Compared to single-pass full-document translation and independent segment-wise translation, our approach achieves significant improvements in BLEU, COMET, and BERTScore across multiple mainstream LLMs. It establishes the first zero-training, highly robust baseline for document-level LLM translation, effectively balancing translation quality, inter-sentence coherence, and inference efficiency.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

LLMs have paved the way for truly simple document-level machine translation, but challenges such as omission errors remain. In this paper, we study a simple method for handling document-level machine translation, by leveraging previous contexts in a multi-turn conversational manner. Specifically, by decomposing documents into segments and iteratively translating them while maintaining previous turns, this method ensures coherent translations without additional training, and can fully re-use the KV cache of previous turns thus minimizing computational overhead. We further propose a `source-primed' method that first provides the whole source document before multi-turn translation. We empirically show this multi-turn method outperforms both translating entire documents in a single turn and translating each segment independently according to multiple automatic metrics in representative LLMs, establishing a strong baseline for document-level translation using LLMs.

Problem

• Addresses omission errors in document-level machine translation.
• Leverages multi-turn conversation for coherent translations without training.
• Proposes source-primed method to improve translation accuracy and efficiency.

Innovation

• Multi-turn conversation for document translation
• Source-primed method enhances translation coherence
• Reuses KV cache to minimize computational overhead